{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2018:3BX3RWIVGF57P5KWXAYGGNSHNC","short_pith_number":"pith:3BX3RWIV","schema_version":"1.0","canonical_sha256":"d86fb8d915317bf7f556b83063364768afff3e9d9dbffc5ee50f0fbf916f165d","source":{"kind":"arxiv","id":"1811.01721","version":1},"attestation_state":"computed","paper":{"title":"Rethinking floating point for deep learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.NA","authors_text":"Jeff Johnson","submitted_at":"2018-11-01T22:13:57Z","abstract_excerpt":"Reducing hardware overhead of neural networks for faster or lower power inference and training is an active area of research. Uniform quantization using integer multiply-add has been thoroughly investigated, which requires learning many quantization parameters, fine-tuning training or other prerequisites. Little effort is made to improve floating point relative to this baseline; it remains energy inefficient, and word size reduction yields drastic loss in needed dynamic range. We improve floating point to be more energy efficient than equivalent bit width integer hardware on a 28 nm ASIC proce"},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"1811.01721","kind":"arxiv","version":1},"metadata":{"license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","primary_cat":"cs.NA","submitted_at":"2018-11-01T22:13:57Z","cross_cats_sorted":["cs.LG"],"title_canon_sha256":"9eed02b805e6798e3b9d476ce563bd33335e92b3b578abebfa9f920bf3504f0c","abstract_canon_sha256":"1742f192aec441bcbd3ea8c0825e6110d630c1ea441777acd74f1a15c22023ca"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-18T00:01:33.457831Z","signature_b64":"aLzsZxpKSrRnGoLfRFwKUjd69bzssWz+n3WdpokL/hQ4RovjWC5MWg9qu25Brfz73BLdIkMvzY8cEz7P0nWQDQ==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"d86fb8d915317bf7f556b83063364768afff3e9d9dbffc5ee50f0fbf916f165d","last_reissued_at":"2026-05-18T00:01:33.457304Z","signature_status":"signed_v1","first_computed_at":"2026-05-18T00:01:33.457304Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Rethinking floating point for deep learning","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"","cross_cats":["cs.LG"],"primary_cat":"cs.NA","authors_text":"Jeff Johnson","submitted_at":"2018-11-01T22:13:57Z","abstract_excerpt":"Reducing hardware overhead of neural networks for faster or lower power inference and training is an active area of research. Uniform quantization using integer multiply-add has been thoroughly investigated, which requires learning many quantization parameters, fine-tuning training or other prerequisites. Little effort is made to improve floating point relative to this baseline; it remains energy inefficient, and word size reduction yields drastic loss in needed dynamic range. We improve floating point to be more energy efficient than equivalent bit width integer hardware on a 28 nm ASIC proce"},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"1811.01721","kind":"arxiv","version":1},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"aliases":[{"alias_kind":"arxiv","alias_value":"1811.01721","created_at":"2026-05-18T00:01:33.457398+00:00"},{"alias_kind":"arxiv_version","alias_value":"1811.01721v1","created_at":"2026-05-18T00:01:33.457398+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.1811.01721","created_at":"2026-05-18T00:01:33.457398+00:00"},{"alias_kind":"pith_short_12","alias_value":"3BX3RWIVGF57","created_at":"2026-05-18T12:32:02.567920+00:00"},{"alias_kind":"pith_short_16","alias_value":"3BX3RWIVGF57P5KW","created_at":"2026-05-18T12:32:02.567920+00:00"},{"alias_kind":"pith_short_8","alias_value":"3BX3RWIV","created_at":"2026-05-18T12:32:02.567920+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":3,"internal_anchor_count":2,"sample":[{"citing_arxiv_id":"1910.09876","citing_title":"Neural Network Training with Approximate Logarithmic Computations","ref_index":25,"is_internal_anchor":true},{"citing_arxiv_id":"2209.05433","citing_title":"FP8 Formats for Deep Learning","ref_index":8,"is_internal_anchor":true},{"citing_arxiv_id":"2604.10484","citing_title":"Strix: Re-thinking NPU Reliability from a System Perspective","ref_index":34,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/3BX3RWIVGF57P5KWXAYGGNSHNC","json":"https://pith.science/pith/3BX3RWIVGF57P5KWXAYGGNSHNC.json","graph_json":"https://pith.science/api/pith-number/3BX3RWIVGF57P5KWXAYGGNSHNC/graph.json","events_json":"https://pith.science/api/pith-number/3BX3RWIVGF57P5KWXAYGGNSHNC/events.json","paper":"https://pith.science/paper/3BX3RWIV"},"agent_actions":{"view_html":"https://pith.science/pith/3BX3RWIVGF57P5KWXAYGGNSHNC","download_json":"https://pith.science/pith/3BX3RWIVGF57P5KWXAYGGNSHNC.json","view_paper":"https://pith.science/paper/3BX3RWIV","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=1811.01721&json=true","fetch_graph":"https://pith.science/api/pith-number/3BX3RWIVGF57P5KWXAYGGNSHNC/graph.json","fetch_events":"https://pith.science/api/pith-number/3BX3RWIVGF57P5KWXAYGGNSHNC/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/3BX3RWIVGF57P5KWXAYGGNSHNC/action/timestamp_anchor","attest_storage":"https://pith.science/pith/3BX3RWIVGF57P5KWXAYGGNSHNC/action/storage_attestation","attest_author":"https://pith.science/pith/3BX3RWIVGF57P5KWXAYGGNSHNC/action/author_attestation","sign_citation":"https://pith.science/pith/3BX3RWIVGF57P5KWXAYGGNSHNC/action/citation_signature","submit_replication":"https://pith.science/pith/3BX3RWIVGF57P5KWXAYGGNSHNC/action/replication_record"}},"created_at":"2026-05-18T00:01:33.457398+00:00","updated_at":"2026-05-18T00:01:33.457398+00:00"}